Some change within the following changelog regressed significantly a couple of DevTools performancec tests: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=e09fbaee1456a373382d2d5bf6ab7321e26e3710&tochange=d231a32316809f8c9efa43ed6f74484ff552c115 After some investigation, it looks like it has been introduced by bug 1437600. Here is a custom try pushing against this bug's patch and its base revision: https://treeherder.mozilla.org/perf.html#/comparesubtest?originalProject=try&originalRevision=f8c7f049f2ce603990c529baeb0f58386bfe02c7&newProject=try&newRevision=37729dba69302c97a79d62efa44dcd4cc6781f8a&originalSignature=f79b6a4f1f8f53326d3056f6c8008c0ff4de0a94&newSignature=f79b6a4f1f8f53326d3056f6c8008c0ff4de0a94&showOnlyImportant=1&framework=1 Here is a set of important regressions reported by this try push: complicated.inspector.open 3.2% complicated.jsdebugger.open 6.5% complicated.jsdebugger.reload 4.5% complicated.netmonitor.open 5.2% complicated.webconsole.open 3.8% ... Given its impact on DAMP, I imagine it may easily have regressed other Talos test suites, but as I don't have access to the bug, I don't know that. Is this a regression we have to live with because of security reasons? Is this regression specific to devtools?

I don't have access to bug 1437600 and can't even make this bug block on it... So ni? you here...

This is expected for investigating a long standing issue which causes more than 2.5k crashes per day. This performance regression should not exists on beta nor release, only in nightly and dev-edition channels.

(In reply to Nicolas B. Pierron [:nbp] {backlog: 39} from comment #2) > This is expected for investigating a long standing issue which causes more > than 2.5k crashes per day. Ah ok, good to know! So if I'm following correctly, that's something we will revert after figuring out the crash? (if yes, is there any bug to track that?) Or will we keep this special mode on nightly/dev-edition to help any further crash investigation?

I did not opened any bug to revert this at the moment, but I would be fine in reusing either this Bug or Bug 1470115 to track this issue. As soon as this investigation is done, we will likely keep this code for fuzzing purposes in Debug builds.

P3 because even though it's an important perf regression, it's temporary and on purpose to allow the investigation of another issue, and won't ship to release users. Having said this, Alex, we probably need to take it into consideration in our damp test results.